International Conference on Electronics, Communication and Aerospace Technology

ICECA 2021

Implementation of an Intelligent Traffic Control System and

Real Time Traffic Statistics Broadcasting

Aman Dubey Akshdeep Sagar Rane

Computer Engineering Information Technology Computer Engineering
Army Institute of Technology, Pune Army Institute of Technology, Pune Army Institute of Technology, Pune
[email protected] [email protected] [email protected]

Abstract— Over the past decades, increasing traffic volume light is empty, traffic control system sitting idle and thus
poses many challenges, dynamic traffic light management is one completely fail to be utilized own resources efficiently. This
of them. Unlike foreign countries, road traffic conditions of India happens because of fixed and same switching time of all the
are heterogeneous due to which detection, counting and traffic lights. In view of above scenario, we can switch to green
classification of vehicles in real-time has become difficult. Due to
fixed and predefined nature of traffic light switching traffic
signal for maximum time where density of traffic is high and
congestion is often. In nutshell, an improved and optimized minimum time where density of traffic is low, then this might
traffic control system is required. Introducing Beagle Bone solve problem.
Black/Rasp Pi to the traffic light system provides numerous
Many attempts were made to transform a traditional traffic
customizations to turn a traditional traffic light into a smart one.
To control traffic at road intersections, system containing
light into smarter one, but these attempts are either sensor-
microcontroller is established on traffic light. Image processing based or manual. Introducing Microcontroller to the traffic
algorithms, such as Haar Cascade and Background Subtraction light helps in applying algorithmic approach to traffic light to
are used to control timer. We have developed traffic control make it smarter in real-time. Image processing algorithm, such
system model for the main intersections of road. The real-time as Haar-cascade and Background subtraction is used for
traffic image, traffic density and other statistics will be sent to dealing with different conditions. These algorithm needs a
server. The data can be broadcasted from server at any time on training period and could be made more accurate. This
demand through digital solutions. The device gives a forecasted approach will allow the traffic timer to change in real-time. In
route to user by contacting different traffic lights on the way of
order to give a real-time view of the upcoming traffic light,
user. The motto of this research is to handle varied traffic
situations efficiently and to save time and money of user.
Region of Influence concept was created, which allows user to
fetch data and statistics from the upcoming traffic light directly
Index Terms— Intelligent Systems, Traffic Control, Traffic and to decide better route. By one click user can receive all the
Lights, Traffic Statistics, Microcontrollers, Real Time Systems statistics and data along with real-time images from the traffic
light, in layman language traffic light is directly
communicating user.
I. INTRODUCTION The paper is organized as follows: section II explains the
Traffic mobbing is a major problem in developed design of the system. Section III discusses the working of the
cities. In this traffic mobbing scenarios peoples are wasting system. Section IV is the result of experimentation. Section V
vehicles fuel and unable to utilized their time [1]. High concludes the paper. Finally, section VI gives future scope of
utilization of present road capacity by managing traffic the project followed by the acknowledgement and key
efficiently is very important. Due to these traffic congestion references used in the research work.
people’s daily schedules getting disturbed as well as some are
going through critical cases [3][5]. If we take example of
ambulance having patient in critical conditions and if there is II. SYSTEM DESIGN
traffic jam on that road, then there are high chances that
ambulance will not be able to reach hospital in time. Due to The work is divided into 4 parts. The first part is to process the
situations like these human lives are in danger. That’s why video signal and image acquisition from fixed camera using
there is need of dynamic intelligent traffic control system who Image Processing. Microcontroller will do the processing part.
can able to handle these traffic conditions efficiently to The second part is to change the timer according to the traffic
circumvent traffic crowding and accidents situations [4][5]. density. The third part is to send all the data to server (in
Current traffic lights control system is ineffective because particular timestamp). Finally, in the last part user can make
signal switching time is predefined and not based on vehicle use of that data according to their needs through an App. The
density. Let’s assume a situation where your vehicle is standing overall block diagram of the proposed system is illustrated
idly before red traffic light, while the lane with green traffic below.

978-1-5090-5686-6/17/$31.00 ©2021 IEEE 33

 International Conference on Electronics, Communication and Aerospace Technology
ICECA 2021

Fig. 1. Proposed Traffic Control System Model

2) Using Background Subtraction, gives the area where
A. Real Time Image Processing the vehicle is not present as shown in Fig.3. According to
the observation, another algorithm can be designed which
A system having camera attached with Microcontroller will change the timer according vehicle density present on
(Beagle Bone Black) is installed on the traffic light. In an that area.
intersection four cameras will be installed and all are
attached to single Microcontroller [10]. Microcontroller can C. Sending Data to the Server
calculate data of an intersection and to control timers of all
Server side script sends Real time image and calculated
traffic lights simultaneously. Camera gives the traffic
statistics to the client side present at database [9]. This data
image to the Microcontroller after certain period according
can be stored according to timestamps, for future use. The
to the algorithm. In microcontroller, Image Processing
data can be further used for security purpose also. Apart
algorithms were there which will process the image to
from the shortest time consuming path that this system is
extract out the needful information in short span of time.
giving, the data will give an extra advantage to the user for
OpenCV (Open Computer Vision) is used for image choosing their path. The microcontroller and the central
processing. Computer vision is a field of informatics, which server are connected through a wireless network. The
teaches computers to see. It is a way computers gather and microcontroller sends all the processed data to the central
interpret visual information from the surrounding server and the central server makes it available on the
environment [1]. While detecting objects Haar taking side mobile app and real time traffic scenario and details can
neighbors as shown in Fig. 2. Job of this algorithm is to add thus be accessed by the user. The data is saved in a
all object pixel intensity and compute difference between database on the central server and can be used at a later
these. This is useful to classify subparts of an image [6]. stage to extract certain important statistics that can be used
Then rectangular windows can be used further to calculate to improve the system further.
count of vehicle density. According to vehicle density
D.Extracting Data from Server through an App
count the traffic light timer can be changed.
Data is stored in database through socket connection. The
B. Changing Timer According to Traffic Density server contains all the data of the traffic light. Data from
different traffic lights are stored in the server. User can
Timer can be changed according to the output of
extract the data from server through any broadcasting
algorithms. Two cases can be made from above algorithms.
system such as App, according to his needs. User can get
1) Using Haar Cascade, gives vehicle density count as the broadcasted data from the server on demand, which
shown in Fig.2. An algorithm can be designed, as per the includes all the statics and other data of the traffic signal.
observation, that cars, trucks, buses took different time to This data can further be used for security purpose. Giving
pass traffic light. Since Haar Cascade is capable enough to all the useful data from traffic signal will allow user to
differentiate between different type of vehicles. analyze the scenario and have a realistic view, without any
physical presence at the traffic signal.

978-1-5090-5686-6/17/$31.00 ©2021 IEEE 34

 International Conference on Electronics, Communication and Aerospace Technology
ICECA 2021

III. WORKING AND EXPERIMENTATION algorithms. These timers are managed and coordinated in such
In this paper, we were used two algorithms to manage a way that waiting time of vehicles will be reduced. If the
traffic signal lights [2]. We have implemented these in vehicle density is at its peak, then timer will be green for
OpenCV: Trained Haar Cascade and Background Subtraction maximum time limit and if there is no traffic then also timer
Algorithms. Haar Cascade uses Haar-classifiers to recognize will be green but for minimum limit (for very less time).
objects as per the algorithm. Through object recognition, total
number of vehicles in-front of the traffic light can be found out.
The count of heterogeneous vehicles such as Bicycle, Scooter,
Bike, Car, Bus, Truck etc. is given to the microprocessor.
According to the count, Microprocessor changes the timer of
the traffic light which will give the dynamic time according to
the traffic density present on lane. Haar Cascade needs to be
trained for accurate results, this training takes some time.
Second Method is Background Subtraction, in which a
reference image is selected in which there were no single
vehicle & matching the real-time images with that reference
image. Traffic lights are controlled based on the percentage
change. For timer count, there will be a maximum and
minimum limit, so that only the dense populated road does not
get whole privilege. Meanwhile, real time image, timer count,
vehicle density, time taken to get out of that traffic light and
other statistics are sent to the server. If the user wants to check
the live status of the traffic light just by sitting at home, then he
is allowed to do so. By the help of this, user have the advantage
to take the minimal time taking path. Fig. 2. Haar Cascade Algorithm Applied on running Video
If the user is headed towards its destination, then during his
journey he can get the live updates by using GPS (global Background Subtraction is also algorithm of OpenCV as
positioning system) technology. For this, a region of influence shown in Fig.3, in which firstly cropping of the unwanted area
is set around a traffic light using longitude and latitude and is done then contours were made of the cars present on the
that region is drawn in such a way that user will have the road. These contours would have different color than that the
choice of taking that path as well as alternative path. When color of background color. A reference image of road without
user enters the region then he will automatically get the real- vehicles was taken and it is used to subtract the background.
time image and statistics of the upcoming traffic light. Along The area in which cars are present on the road are subtracted
with that user will be provided with the less time taking path, from empty road. Remaining area is the area of the empty
so that traffic can be diverted to other empty traffic light also. road, as shown in Fig. From this algorithm that area can be
With the help of Microcontroller, Image processing, GPS and extracted out where the vehicle is not present and the timer
Android App, traffic problem at the intersection can be solved can be changed accordingly. In Fig. 2, Haar Cascade
out. Microcontroller will do the processing part. The second Algorithm is used, which can be made more accurate by
part is to change the timer according to the traffic density. The training Haar classifier. This classifier is used to make the
third part is to send all the data to server timestamp. Finally, in algorithm more perfect.
the last part user can make use of that data according to their
needs through an App. The overall block diagram of the
proposed system is illustrated below.

To implement proposed traffic control system, we have used

Beagle Bone Black development kit with OpenCV and
Adafruit python libraries. This system mainly consists of two
parts i.e., Hardware part and Software part. In hardware part, a
Camera is attached to Beagle Bone Black which is connected
to the server through an unguided medium. The camera will
feed the Beagle Bone Black with real-time image. Beagle
Bone’s responsibility is to process the image and to give
useful output from the image. From image one can extract out
many details like time taken by vehicles to pass the signal,
waiting time of vehicles, route which is more suitable, cars
present on the lane, empty area of the lane etc. Management of Fig. 3. Background Subtraction Algorithm (OpenCV)
traffic and coordination of traffic lights (in case of
intersection) is possible because of these results. Traffic light
timers can be changed according to the output of OpenCV

978-1-5090-5686-6/17/$31.00 ©2021 IEEE 35

 International Conference on Electronics, Communication and Aerospace Technology
ICECA 2021

Software part consists of Server, databases, and

Broadcasting Device (Android App). The data is given to the
server by microcontroller using IoT (Internet of Things)
protocol. Data is stored in server according to the timestamp.
User can get the traffic status of the upcoming traffic light or
the traffic lights which user is going to pass through out his
journey.

IV. RESULTS
If user selects any path then a map is given to the user,
which contains red marks, where this system is installed. If the
user clicks on any of the red mark, then whole details and Fig. 5. Average Waiting Time of Vehicles in Front of Traffic Light
statistics of the traffic light will be given to the user as shown
in Fig.4. The area around traffic light is called the ‘REGION The graph is drawn by experimental calculations of traffic
OF INFLUENCE’ as shown by the encircled area in Fig.4. light of Pune. This is the average time wasted by vehicles
When the user enters the influence region, a notification will because of inefficient traffic management system. Using above
pop up into the user’s mobile which shows the real-time image algorithms can make easy and efficient management of traffic
and other statistics of the upcoming traffic light. Along with at intersection. From Fig.5, another result can be deducted that
that, most suitable and less time taking path will be given to the peak hours are from 8AM to 11AM and 6PM to 8PM, where
user. This region is set according to user’s requirement, so that the waiting time of vehicles is the most and this is the time
user will have the option to either take that path or to choose where vehicular density is high. So, to make system well
another path. Suitable path will be decided by taking all the organized, traffic needs to be maintained efficiently.
other traffic lights into consideration coming to user’s path.
This GPS system will be established by using longitude and TABLE I. COMPOSITION OF TRAFFIC
latitude of the traffic light. By this way traffic light can directly
VEHICLE TYPE Percentage
communicate the end user and can send and receive data. This
part is an add-on to the project. 2-wheeler 53%
Cycle 1%
Auto 11%
Car 22%
Bus 9%
Trucks 4%

Table I, shows the percentage of vehicles during peak hour.

Major composition of traffic consists of 2-wheelers and cars
which comprises almost 75% of the total traffic. Above table
also shows that traffic in India is highly heterogeneous. These
experiments are responsible for making Haar Cascade classifier
more efficient. As new data is generated over time more
statistic details are available to train Haar Cascade algorithm
and thus the algorithm provides more accurate results and
Fig. 4. Shortest Path with Region of Influence
consequently performance of the system improves.
V. CONCLUSION
Following are the experimentation results of the
Vishrantwadi, Pune Traffic light. By considering only one side Traffic congestion on Indian roads is one of the big issues.
of the lane average waiting time and maximum vehicle density Our current work is a systematic solution to handle this issue.
are few outcomes derived from that experiment. We have By using this unique research work, Image processing shown
collected images with the help of camera from 06.00 AM this technique is an efficient to switch the states of traffic
morning to 11.59 PM midnight. This simulation was done for lights. Our proposed traffic control system proved that it can
around 18 hours and made remarks on graph as shown in Fig. 5 minimize traffic mobbing and reduces waiting time of vehicles
about average waiting time verses average one hour period. In in front of traffic signal. To make real time analysis of traffic in
this case, we assumed every lane contains predefined and fixed modern cities, we configured and coupled actual traffic images
duration of red light signal. taken with a microcontroller. This projected solution
guarantees that the Avg. waiting time of vehicles in front of
Fig.5 shows the Average Waiting time of Vehicles in front of traffic signal will be lesser than present traffic control systems.
traffic light throughout the day from 6 AM to 12 AM.

978-1-5090-5686-6/17/$31.00 ©2021 IEEE 36

 International Conference on Electronics, Communication and Aerospace Technology
ICECA 2021

Moreover, this processed data is made available to the user [2] Vikramaditya Dangi, Amol Parab, Kshitij Pawar & S.S Rathod, “Image
Processing Based Intelligent Traffic Controller”, Undergraduate
through App which equips user with various paths that are less Academic Research Journal (UARJ), Vol.1, Issue 1, 2012.
time consuming. The data available to the user is more accurate [3] Papageorgiou M., Diakaki C., Dinopoulou V., Kotsialos, A.,“Review of
than other companies’ data, since this method is more towards road traffic control strategies”, Proceedings of IEEE, Vol. 91, Issue 12,
algorithmic approach. By this way, real-time scenarios can be pp. 2043-2067, November 2004.
created for user even if user is not present at the place. This [4] Georgios Vigos, Markos Papageorgioua, Yibing Wangb, “Real-time
system is capable of handling multiple tasks, so no need of estimation of vehicle-count within signalized links”, Journal of
Transportation Research Part C: Emerging Technologies, Volume 16,
deploying different systems for different purposes. Since ‘time Issue 1, pp.18–35, February 2008.
is money’, the more time you save, more money you can earn. [5] Michael W. Szeto and Denos C. Gazis, “Application of Kalman
User’s time can be saved if user chooses the path given by Filtering to the Surveillance and Control of Traffic Systems”, Journal of
App. Smart timer implementation saves fuel since user do not Transportation Science, vol. 6 pp.. 4419-439 , November 1972.
have to sit idly waiting for traffic light to turn green. [6] Celil Ozkurt and Fatih Camci,“ Automatic Traffic Density Estimation
Algorithmic approach opted in this project is more effective as and Vehicle Classification For Traffic Survillance Systems Using Neural
Networks”, Journal of Mathematical and Computational Applications,
compared to previous attempts on making the smart traffic Vol. 14, No. 3, pp. 187-196, 2009.
system which uses sensors. This system overcomes the [7] D. Beymer, P. McLauchlan, B. coifman and J. Malik, "A real-time
disadvantages of sensors, since sensors are fragile and need computer vision system for measuring traffic parameters", IEEE conf.on
maintenance. Emergency vehicles can also reach to the computer vision and pattern Recognition, (1997), pp. 495-501.
destination before it’s too late. The saved data can be further [8] Pratishtha Gupta, G.N Purohit, Sweta Pandey, “Traffic Load
Computation for Real Time Traffic Signal Control”, International
used for any security purpose. At last we can conclude that our Journal of Engineering and Advanced Technology, Vol. 2, Issue 4, April
proposed system is very much efficient which is capable of 2013.
handling multiple tasks in real-time. [9] B. Jin, D. Zou and Y. Gan, "Research and design of traffic detection
based on GPS", IEEE conf. on advanced computer control (ICACC),
(2010) March.
VI. FUTURE SCOPE [10] “BeagleBone Black System Reference Manual”, Revision A5.2April
11, 2013. (http://www.adafruit.com).
[11] Celil Ozkurt, et al. Automatic Traffic Density Estimation And Vehicle
This system can be further extended by making it Classification For Traffic Surveillance Systems Using Neural Networks.
community based service. It is basically driven by a Mathematical and Computational Applications, Vol. 14, No. 3, pp. 187-
community of people who are currently facing the situation. If 196, 2009.
any incident happens, then the user will update about it in the [12] Wisnu Jatmiko, et.al,” Detection and Counting of Vehicles Based on
Video Processing In Distributed Traffic System”, International
App, so that other users would take care of this and can avoid Conference on Advance Computer Science and Information System
that route. Accordingly, the App can be made more efficient. 2010.
Emergency vehicles such as Ambulances, Fire Brigades can be [13] Jie Zhou.et al. Moving Vehicle Detection for Automatic Traffic
provided with special privileges. They can directly contact the Monitoring. IEEE Transactions on Vehicular Technology, Vol. 56, No.
server room, so that traffic light coming in their routes can be 1, 2007.
managed accordingly. Traffic lights can be made green for its
maximum time limit, till the vehicles passes. By designing
appropriate algorithms, exact time between source and
destination can be found out with the help of real time speeds
of vehicle since we have the exact time of vehicle to pass
through that traffic light. Moreover, this system can also be
used by government, local bodies and police authorities for
security purpose or for any estimation.

VII. ACKNOWLEDGMENT
Authors want to give special thanks to HoD Computer
Department, Army Institute of Technology, Pune for their
enormous support and help. All the facilities and financial
support provided by the institution for the research work are
acknowledged.

VIII. REFERENCES
[1] Sabya sanchi kanojia, “Real –time Traffic light control and Congestion
avoidance system”, International Journal of Engineering Research and
Applications (IJERA), pp.925-929, Vol. 2, Issue 2,Mar-Apr 2012.

978-1-5090-5686-6/17/$31.00 ©2021 IEEE 37